Search papers, labs, and topics across Lattice.
This paper surveys the emerging field of post-training techniques for end-to-end autonomous driving models, which directly map multimodal inputs to future driving trajectories. It highlights the limitations of traditional imitation learning in safety-critical environments, where small errors can lead to significant failures, and long-horizon objectives like safety and comfort are not adequately captured. By categorizing existing literature into four families based on supervision types, the authors provide a comprehensive framework to understand and advance post-training methods in this domain.
Post-training techniques could be the key to overcoming the limitations of traditional imitation learning in autonomous driving, ensuring safer and more reliable vehicle behavior in complex environments.
End-to-end models that map multimodal inputs directly to future trajectories/maneuvers have emerged as an increasingly prominent research paradigm in autonomous driving. This class of models includes both Vision-Language-Action models and trajectory-generative planners. Unlike classic machine learning applications, autonomous vehicles operate in safety-critical and interaction-intensive environments where traditional open-loop imitation of expert demonstrations is not sufficient to ensure reliability. In particular, small execution errors can accumulate over time, while recovery behaviors are scarce in training data. In addition, long-horizon objectives such as safety and driving comfort are not captured by pointwise labels either. These limitations have motivated a shift toward post-training techniques, which further refine driving policies beyond pure imitation. This survey presents a unified view of post-training for autonomous driving by defining its scope and organizing the existing literature into four major families based on the form of supervision they use. For each family, we discuss its capabilities, limitations, and open challenges. We aim to facilitate a systematic understanding of this emerging area and stimulate future research on reliable and efficient post-training for autonomous driving.
